Radiacion por fuego
-
Upload
wil-vasquez-c -
Category
Documents
-
view
214 -
download
0
Transcript of Radiacion por fuego
-
8/19/2019 Radiacion por fuego
1/146
Evaluation of Thermal Radiation Models
for Fire Spread Between Objects
by
Rob Fleury
Supervised by
Dr. Michael Spearpoint and
Associate Professor Charles Fleischmann
2010
A thesis submitted in partial fulfilment of the requirements for the
degree of Master of Engineering in Fire Engineering
Department of Civil and Natural Resources Engineering
University of Canterbury
Christchurch, New Zealand
-
8/19/2019 Radiacion por fuego
2/146
-
8/19/2019 Radiacion por fuego
3/146
i
ABSTRACT
Fire spread between objects within a compartment is primarily due to the impingement of
thermal radiation from the fire source. In order to estimate if or when a remote object
from the fire will ignite, one must be able to quantify the radiative heat flux being received
by the target. There are a variety of methods presented in the literature that attempt to
calculate the thermal radiation to a target; each one based on assumptions about the fire.
The performance of six of these methods, of varying complexity, is investigated in this
research. This includes the common point source model, three different cylindrical
models, a basic correlation and a planar model. In order to determine the performance of
each method, the predictions made by the models were compared with actual
measurements of radiant heat flux. This involved taking heat flux readings at numerous
locations surrounding a propane gas burner. Different fire scenarios were represented by
varying the burner geometry and heat release rate. Video recordings of the experiments
were used to determine the mean flame heights using video image analysis software.
After comparing the measured data with predictions made by the theoretical radiation
methods, the point source model was found to be the best performing method on average.
This was unexpected given the relative simplicity of the model in comparison to some ofits counterparts. Additionally, the point source model proved to be the most robust of the
six methods investigated, being least affected by the experimental variables. The Dayan
and Tien method, one of the cylindrical models, was the second most accurate over the
range of conditions tested in this work.
Based on these findings, recommendations are made as to the most appropriate method for
use in a radiation sub-model within an existing zone model software. The accuracy shown
by the point source model, coupled with its ease of implementation, means that it shouldbe suitable for such a use.
-
8/19/2019 Radiacion por fuego
4/146
ii
-
8/19/2019 Radiacion por fuego
5/146
iii
ACKNOWLEDGEMENTS
The author would like to offer sincere thanks to the following people and organisations
who have assisted with this research, including:
• Supervisors Mike Spearpoint and Charley Fleischman for generously giving their
time, knowledge and guidance
• The Foundation for Research, Science and Technology for providing funding for
the project
• Building Research Association of New Zealand Ltd for providing the project itself,
with special thanks to Greg Baker, Colleen Wade and Amanda Robbins of
BRANZ Ltd for their support• The University of Canterbury and the Fire Engineering programme
• Grant Dunlop and Bob Wilsea-Smith whose expertise in the laboratory was
invaluable
• Roger Nokes for the use of the ImageStream software and assistance in applying it
• Finally, the New Zealand Fire Service Commission for their continued support of
the Fire Engineering Programme at the University of Canterbury
-
8/19/2019 Radiacion por fuego
6/146
-
8/19/2019 Radiacion por fuego
7/146
v
3.4.3 Calorimetry ............................................................................................... - 38 -
3.4.4 Video equipment ....................................................................................... - 39 -
3.4.5 Load cell ................................................................................................... - 39 -
3.4.6 Mass flow controllers................................................................................ - 40 -
3.5 Experimental Procedure ............................................................................. - 41 -3.6 Experimental Data Analysis Technique...................................................... - 42 -
3.6.1 Heat release rate ........................................................................................ - 42 -
3.6.2 Radiant heat flux ....................................................................................... - 43 -
3.7 Flame Height Determination ...................................................................... - 45 -
3.7.1 Video image analysis................................................................................. - 46 -
EXPERIMENTAL RESULTS AND DISCUSSION .............................................. - 53 -
4.1 Heat Release Rate ...................................................................................... - 53 -
4.2 Radiant Heat Flux ...................................................................................... - 55 -
4.2.1 Variation of heat flux with distance from fire ............................................ - 55 -
4.2.2 Variation of heat flux with heat release rate............. ........... ......... .......... .... - 56 -
4.2.3 Comparison between front and side gauges ............................................... - 57 -
4.2.4 Variation of heat flux with height above fire...... ........... ......... ......... ........... - 59 -
4.2.5 Comparison between burner aspect ratios .......... ........... ......... ......... ........... - 60 -
4.2.6 Variation of heat flux with burner angle .................................................... - 62 -
4.2.7 Comparison between central and offset gauges............. ......... ......... ........... - 63 -
4.2.8 Comparison between vertical and horizontal gauges........... ......... .......... .... - 66 -
4.2.9 Repeatability of results .............................................................................. - 68 -
4.3 Flame Height.............................................................................................. - 69 -
4.3.1 Comparison with correlations .................................................................... - 71 -
4.3.2 Buoyancy driven flame validation ............................................................. - 72 -
THEORETICAL MODEL ANALYSIS.................................................................. - 75 -
5.1 Variables and Constants used in Radiation Models.......... .......... .......... ....... - 75 -
5.2 Overview of Model Results ........................................................................ - 76 -
5.3 Basic Models.............................................................................................. - 78 -
5.4 Cylindrical Models..................................................................................... - 86 -
5.5 Planar Model.............................................................................................. - 92 -
5.6 Sensitivity to Inputs.................................................................................... - 97 -
5.6.1 Radiative fraction ...................................................................................... - 97 -
5.6.2 Effective absorption coefficient ................................................................. - 99 -
-
8/19/2019 Radiacion por fuego
8/146
vi
5.6.3 Flame temperature................................................................................... - 101 -
5.6.4 Flame height............................................................................................ - 102 -
5.6.5 Effective emissive power......................................................................... - 104 -
5.6.6 Distance convention ................................................................................ - 104 -
5.7 Summary of Models ................................................................................. - 106 -5.8 Limitations to Results............................................................................... - 110 -
5.9 Recommendation for BRANZFIRE Radiation Sub-Model............. ......... .. - 111 -
CONCLUSIONS.................................................................................................... - 114 -
6.1 Experimental Results................................................................................ - 114 -
6.1.1 Radiant heat flux ..................................................................................... - 114 -
6.1.2 Flame height............................................................................................ - 115 -
6.2 Performance of Models ............................................................................ - 115 -
6.2.1 Recommendation..................................................................................... - 116 -
6.3 Further Research ...................................................................................... - 117 -
REFERENCES ...................................................................................................... - 119 -
APPENDIX A – EMISSIVITY PLOTS................................................................ - 123 -
APPENDIX B – CONTOUR PLOTS.................................................................... - 125 -
APPENDIX C – HORIZONTAL RESULTS........................................................ - 129 -
-
8/19/2019 Radiacion por fuego
9/146
vii
LIST OF FIGURES
Figure 1: Overview of development scope for risk-informed fire design tool(BRANZ Ltd, 2007). The research outlined in this thesis comprisespart of the highlighted section, Develop item-item fire spread model .......... - 5 -
Figure 2: (a) Schematic of point source model (Karlsson & Quintiere, 2000), (b)Schematic and notation for point source model (Beyler, 2002) .......... ....... - 15 -
Figure 3: Cylindrical flame shape configuration factor geometry for vertical andhorizontal targets at ground level (Beyler, 2002)............ .......... .......... ....... - 17 -
Figure 4: Two-cylinder representations of the configuration factor for target aboveground level (Beyler, 2002) ...................................................................... - 17 -
Figure 5: Schematic of radiation exchange between a target element, dA, and ahomogeneous cylindrical flame (Karlsson & Quintiere, 2000) ......... ......... - 22 -
Figure 6: (a) The rectangular planar model is made up of two perpendicularintersecting planes, (b) The planes intersect at the centre of the fire,which can be approximated as a rectangular cuboid..... ......... ......... ........... - 24 -
Figure 7: Pictorial representation and notation for the configuration factor from afinite rectangle to a differential element (Howell, 2008) .......... .......... ....... - 26 -
Figure 8: Example situation where the normal of the rectangle to the target lieswithin the bounds of the fire (area A2). The fire must be divided intofour individual rectangles and the configuration factors for each addedto achieve the overall configuration factor ................................................ - 27 -
Figure 9: Example situation where the normal of the rectangle to the target liesoutside of the bounds of the fire (area A2). The overall configurationfactor is found by adding factors F 1 and F 2 then subtracting factors F 3 and F 4....................................................................................................... - 27 -
Figure 10: Gas delivery pipe within the gas burners (not to scale) ........ ........... ........... - 33 -
Figure 11: The 3:1 aspect ratio burner, filled with fired clay balls to aid indiffusion................................................................................................... - 33 -
Figure 12: Relative angles of burner positions used in experiments. 2:1 burneraspect ratio depicted ................................................................................. - 34 -
Figure 13: Schematic of laboratory layout (not to scale)........... ........... .......... ......... .... - 35 -
Figure 14: Photograph taken during experimental work, showing heat flux gaugetrolleys located at 1.0 m from burner centre. Burner shown has 1:1aspect ratio ............................................................................................... - 35 -
Figure 15: Side view of heat flux gauge trolley on its tracks................. ........... ........... - 37 -
Figure 16: Front view of heat flux gauge trolley with offset and central gaugepositions indicated.................................................................................... - 37 -
Figure 17: Example of raw heat flux gauge data, from side gauges of Test 13...... ..... - 43 -
-
8/19/2019 Radiacion por fuego
10/146
viii
Figure 18: First 18 minutes of Test 13, corresponding to the 300 kW heat releaserate ........................................................................................................... - 44 -
Figure 19: Distribution of raw heat flux data for Test 11, 300 kW, 0.5 m from firecentre, 0.0 m above flame base ................................................................. - 45 -
Figure 20: Definition of mean flame height, H, from measurements of flame
intermittency, I ......................................................................................... - 46 -Figure 21: Process of image manipulation using ImageStream by applying filters.
(a) raw image from single frame of video, (b) extraction of red colourintensity, (c) colour removed below certain red colour threshold, (d)amplification of red colour intensity, (e) intensities converted andnormalised to real Boolean values ............................................................ - 48 -
Figure 22: Time averaged contour plot for the 2:1 burner at 300 kW. Vertical axisgives flame height (in mm). Scale on right hand side givesprobabilities of flames existing at different locations .......... ........... ........... - 49 -
Figure 23: Alternative forms of ImageStream output for 2:1 burner at 300 kW .......... - 50 -
Figure 24: Time-averaged contour plots of the 2:1 burner at 300 kW resulting from(a) 10 mm grid, (b) 20 mm grid, (c) 40 mm grid and (d) 80 mm grid ........ - 51 -
Figure 25: Comparison of heat release rates (from Test 2). Shown are desiredvalues from mass flow controllers, calculated values based on propanemass loss and calculated values based on ODC (30 second movingaverage plotted)........................................................................................ - 54 -
Figure 26: Radiant heat flux vs horizontal distance from fire centre. From Test 11,2:1 burner, front gauges, gauge height above flame base = 0.5 m.............. - 55 -
Figure 27: Radiant heat flux vs heat release rate of fire. From Test 11. Frontgauges, distance from fire centre = 0.5 m.............. ......... .......... .......... ....... - 57 -
Figure 28: Typical view of a 300 kW fire from the 3:1 aspect ratio burner whenviewed from (a) the side and (b) the front ................................................. - 58 -
Figure 29: Radiant heat flux vs horizontal distance from fire centre. From Test 9,3:1 burner, gauge height above flame base = 0.5 m............ ......... .......... .... - 59 -
Figure 30: Radiant heat flux vs height above base of flame. From Test 11, 2:1burner, front gauges, horizontal distance from fire centre = 0.5 m........ ..... - 60 -
Figure 31: Schematic of radiation from a cylindrical fire to a target located at (a)the base height of the flame and (b) the mid-height of the flame. Thethickness of the arrows represent the amount of radiation beingreceived from different areas of the fire (not to scale)......... ........... ........... - 60 -
Figure 32: Radiant heat flux vs horizontal distance from fire centre. From Tests 7,9 and 11. Heat release rate = 300 kW, front gauges, gauge heightabove flame base = 0.5 m ......................................................................... - 61 -
Figure 33: Radiant heat flux vs heat release rate for (a) side gauges and (b) frontgauges. From Tests 7, 9 and 11. Horizontal distance from fire centre =0.5 m, gauge height above flame base = 0.5 m.......... ......... ............ ........... - 62 -
Figure 34: Comparison of radiative heat flux measurements from burners orientedat 0° and 45°. Data is presented from all burner aspect ratios, gauge
-
8/19/2019 Radiacion por fuego
11/146
ix
positions, heat release rates, distances from fire and heights aboveflame base. From tests 1-12. .................................................................... - 63 -
Figure 35: Radiant heat flux vs horizontal distance from fire centre, comparingcentral and offset gauge positions. From Tests 3, 5, 7 and 9. Heatrelease rate = 300 kW, front gauges, gauge height above flame base =0.5 m ........................................................................................................ - 65 -
Figure 36: Comparison of radiative heat flux measurements from central and offsetgauges. Data is presented from all burner aspect ratios, burner angles,heat release rates, distances from fire and heights above flame base.From tests 1-12......................................................................................... - 65 -
Figure 37: Radiant heat flux vs horizontal distance from fire centre. From Tests 11and 16, 2:1 burner, front gauges, gauge height above flame base = 0.5m.............................................................................................................. - 66 -
Figure 38: Radiant heat flux vs height above base of flame. From Test 16, 2:1burner, front gauges, horizontal distance from fire centre = 0.5 m........ ..... - 67 -
Figure 39: Radiant heat flux vs horizontal distance from fire centre. From Tests14-16. Heat release rate = 300 kW, front gauges, gauge height aboveflame base = 0.5 m ................................................................................... - 68 -
Figure 40: Comparison of radiative heat flux measurements from Tests 9 and 13(replicate tests) ......................................................................................... - 69 -
Figure 41: Mean flame heights vs heat release rate for all three burner aspectratios, generated by ImageStream ............................................................. - 70 -
Figure 42: Flame intermittency vs normalised flame height for all burner aspectratios and heat release rates tested, generated by ImageStream ................. - 71 -
Figure 43: Mean flame height vs heat release rate for the 2:1 burner as determinedusing ImageStream , the Heskestad correlation and the Thomascorrelation................................................................................................ - 72 -
Figure 44: Mean flame height normalised by source diameter vs non-dimensionalheat release rate for large number of independent experiments (adaptedfrom Heskestad (2002)). Solid symbols represent experimental datacollected as part of this research ............................................................... - 74 -
Figure 45: Mean flame height normalised by source diameter vs non-dimensionalheat release rate for different burner aspect ratios ........ ........... ......... ......... - 74 -
Figure 46: Radiant heat flux vs horizontal distance from fire centre. Comparisonof experimental results with all models. From Test 11, 2:1 burner,300 kW, gauge height above flame base = 0.5 m ......... ........... ......... ......... - 76 -
Figure 47: Radiant heat flux vs height above flame base. Comparison ofexperimental results with all models. From Test 11, 2:1 burner, 300kW, gauge distance from fire centre = 0.5 m........... ........... ........... .......... .. - 77 -
Figure 48: Radiant heat flux vs heat release rate of fire. Comparison ofexperimental results with all models. From Test 11, 2:1 burner, gaugedistance from fire centre = 0.5 m, gauge height above flame base =0.5 m ........................................................................................................ - 77 -
-
8/19/2019 Radiacion por fuego
12/146
x
Figure 49: Radiant heat flux vs horizontal distance from fire centre. Comparisonof experimental results with basic models. From Test 11, 2:1 burner,300 kW, gauge height above flame base = 0.5 m .......... .......... ......... ......... - 79 -
Figure 50: Radiant heat flux vs height above flame base. Comparison ofexperimental results with basic models. From Test 11, 2:1 burner,distance from fire centre = 0.5 m, at heat release rate of (a) 300 kW and(b) 100 kW............................................................................................... - 80 -
Figure 51: Comparison of measured and predicted radiative heat flux using (a)Shokri & Beyler correlation and (b) point source model. Data takenfrom Tests 7, 9 and 11 .............................................................................. - 81 -
Figure 52: Radiant heat flux vs horizontal distance from fire centre for horizontaltargets. From Test 16, 2:1 burner, 300 kW, gauge height above flamebase = 0.0 m ............................................................................................. - 83 -
Figure 53: Comparison of measured and predicted radiative heat flux to horizontaltargets using point source model. Data taken from Tests 14-16 ........... ..... - 84 -
Figure 54: Schematic of thermal radiation from point source to horizontal targets...... - 85 -
Figure 55: Radiant heat flux vs horizontal distance from fire centre. Comparisonof experimental results with cylindrical models. From Test 11, 2:1burner, 300 kW, gauge height above flame base = 0.5 m ......... .......... ....... - 87 -
Figure 56: Radiant heat flux vs height above flame base. Comparison ofexperimental results with cylindrical models. From Test 11, 2:1 burner,gauge height above flame base = 0.5 m, at heat release rate of (a) 300kW and (b) 100 kW .................................................................................. - 87 -
Figure 57: Comparison of measured and predicted radiative heat flux using (a)Shokri & Beyler detailed method, (b) Mudan method and (c) Dayan &Tien method. Data taken from Tests 7, 9 and 11. ......... .......... ......... ......... - 89 -
Figure 58: Comparison of measured and predicted radiative heat flux using Dayan& Tien method, including linear trend lines through data series. Datataken from Tests 7, 9 and 11. .................................................................... - 90 -
Figure 59: Radiant heat flux vs horizontal distance from fire centre for horizontaltargets. From Test 16, 2:1 burner, 300 kW, gauge height above flamebase = 0.0 m ............................................................................................. - 92 -
Figure 60: Radiant heat flux vs horizontal distance from fire centre. Comparisonof experimental results with rectangular planar model. From Test 11,2:1 burner, 300 kW, gauge height above flame base = 0.5 m ......... ........... - 93 -
Figure 61: Radiant heat flux vs height above flame base. Comparison of
experimental results with planar model. From Test 11, 2:1 burner,gauge height above flame base = 0.5 m, at heat release rate of (a) 300kW and (b) 100 kW .................................................................................. - 94 -
Figure 62: Comparison of measured and predicted radiative heat flux usingrectangular planar model. Data taken from Tests 7, 9 and 11 ........ ........... - 95 -
Figure 63: Radiant heat flux vs horizontal distance from fire centre for horizontaltargets. From Test 16, 2:1 burner, 300 kW, gauge height above flamebase = 0.0 m ............................................................................................. - 96 -
-
8/19/2019 Radiacion por fuego
13/146
xi
Figure 64: Radiant heat flux vs horizontal distance from fire centre. Comparisonof results using different radiative fractions for point source model.From Test 11, 2:1 burner, 300 kW, gauge height above flame base =0.5 m ........................................................................................................ - 98 -
Figure 65: Average percentage error from experimental results vs radiativefraction used in calculation of predictions by point source model. FromTest 11 (all data)....................................................................................... - 99 -
Figure 66: Radiant heat flux vs horizontal distance from fire centre. Comparisonof results using different effective absorption coefficients for (a) Dayan& Tien method and (b) rectangular planar model (front gauges). FromTest 11, 2:1 burner, 300 kW, gauge height above flame base = 0.5 m ..... - 100 -
Figure 67: Average percentage error from experimental results vs effectiveabsorption coefficient used in calculation of predictions for Dayan &Tien method and rectangular planar model. From Test 11 (all data)....... - 101 -
Figure 68: Average percentage error from experimental results vs flametemperature used in calculation of predictions for Dayan & Tien
method and rectangular planar model. From Test 11 (all data) .......... ..... - 102 -Figure 69: Comparison of average percentage errors from experimental data using
two different methods of flame height determination: ImageStream andrecommended correlations. Data from Tests 1-16 ......... .......... .......... ..... - 103 -
Figure 70: Comparison of average percentage errors from experimental results atdifferent heat release rates for different conventions of measuringdistance between target and fire. From rectangular planar model, Test11, front and side gauges, height above flame base = 0.5 m........... ......... - 105 -
Figure A 1: Total emissivity of water-vapour in a mixture of total pressure of
1 atm (Beyler, 2002)............................................................................... - 123 -Figure A 2: Total emissivity of carbon dioxide in a mixture of total pressure of
1 atm (Beyler, 2002)............................................................................... - 124 -
Figure B 1: ImageStream contour plots to determine flame height for 1:1 burner at(a) 100 kW, (b) 150 kW, (c) 200 kW, (d) 250 kW and (e) 300 kW.......... - 126 -
Figure B 2: ImageStream contour plots to determine flame height for 2:1 burner at(a) 100 kW, (b) 150 kW, (c) 200 kW, (d) 250 kW and (e) 300 kW.......... - 127 -
Figure B 3: ImageStream contour plots to determine flame height for 3:1 burner at(a) 100 kW, (b) 150 kW, (c) 200 kW, (d) 250 kW and (e) 300 kW.......... - 128 -
Figure C 1: Measured vs predicted heat fluxes for horizontal targets for (a) pointsource model, (b) Shokri & Beyler detailed method, (c) Mudan method,(d) Dayan & Tien method and (e) rectangular planar model.......... .......... - 130 -
-
8/19/2019 Radiacion por fuego
14/146
xii
LIST OF TABLES
Table 1: Gauge and burner settings for each experimental test ........ .......... .......... ....... - 41 -
Table 2: Results of flame height validation ................................................................ - 51 -Table 3: Coefficient of variation data for Test 11, front gauges, gauge height above
flame base = 0.5 m ................................................................................... - 56 -
Table 4: Summary of percentage errors for point source model. Data from Tests 7,9 and 11.................................................................................................... - 82 -
Table 5: Summary of percentage errors for basic models. Breakdown of fire aspectratios. Data from Tests 1-13...................................................................... - 83 -
Table 6: Summary of percentage errors for cylindrical models. Breakdown of fireaspect ratios. Vertical targets only (Tests 1-13)....... ........... ......... .......... .... - 90 -
Table 7: Summary of percentage errors for Dayan & Tien method. Data from Tests7, 9 and 11................................................................................................ - 91 -
Table 8: Summary of average absolute percentage errors for rectangular planarmodel. Breakdown of fire aspect ratios. Data from Tests 1-13 ......... ......... - 95 -
Table 9: Summary of average absolute percentage errors from experimental resultsfor all theoretical models ........................................................................ - 106 -
Table 10: Summary of results for different target orientations............. ......... .......... .. - 107 -
Table 11: Summary of results for different burner aspect ratios............ ........... ......... - 107 -
Table 12: Summary of results for different radiant heat flux ranges. From Tests 9,11 and 13................................................................................................ - 108 -
Table 13: Summary of results for different target positions ........... ......... ......... ......... - 108 -Table 14: Summary of results for different burner angles......... ........... ......... .......... .. - 108 -
-
8/19/2019 Radiacion por fuego
15/146
xiii
NOMENCLATURE
A Fire/pool surface area (cross-sectional) (m²)
A2 Finite rectangle area (rectangular planar model) (m²)
dA1 Differential target element ( - )
c p Specific heat of air at constant pressure (kJ/kgK)
cv Coefficient of variation (%)
D Fire/pool diameter (m)
E (Effective) emissive power of flame (kW/m²)
E max Equivalent black body emissive power (kW/m²)
E s Emissive power of smoke (kW/m²)
F 12
Configuration/shape/view factor from fire to target ( - )
F 12,max Maximum configuration factor at a point ( - )
F Configuration factor from Plane to target ( - )
F Configuration factor from Plane to target ( - )
g Gravitational acceleration (9.81 m/s²)
H Flame height (m)
H T Height of target relative to height of equivalent point
source at H /2 (m)
H c Heat of combustion (kJ/kg) L Distance of target from centre of fire/pool (m)
Path length from flame surface to receiving target (m)
l f Length of flame (rectangular planar model) (m)
m f Final mass of gas bottles (kg)
mi Initial mass of gas bottles (kg)
∞′′m Mass burning rate per unit area (kg/m²s)
n Unit normal vector to differential target element ( - )
p' w Partial pressure of water vapour (atm)
pw Partial pressure path length parameter (atm m)
Q Heat release rate of fire (kW)
*Q Non-dimensional heat release rate ( - )
r Q Radiative energy output of fire (kW)
-
8/19/2019 Radiacion por fuego
16/146
xiv
q ′′ Radiant heat flux (kW/m²)
R Distance from point source to target (m)
r Fire/pool radius (m)
RH Relative humidity (%)
s Extinction coefficient (m -1)
stdev Standard deviation of sample (any)
t Time (s)
T a Ambient temperature (K)
T f Flame temperature (K)
u Component of n in i direction ( - )
v Component of n in j direction ( - )
w Component of n in k direction ( - )
w f Width of flame (rectangular planar model) (m)
x Sample mean (any)
x Position of target relative to origin in i direction (m)
y Position of target relative to origin in j direction (m)
z Position of target relative to origin in k direction (m)
Greek Symbols
c Carbon dioxide absorption coefficient ( - )
w Water vapour absorption coefficient ( - )
Mean value of in Dayan & Tien method (radians)
Emissivity ( - )
c Carbon dioxide emissivity ( - )
w Water vapour emissivity ( - )
Angle between normal to target and line of sight from
target to point source location (radians)
0 Angle between z axis and line of sight from target to
centre-top of cylinder, Dayan & Tien method (radians) Effective flame absorption coefficient (m -1)
a Ambient air density (kg/m³)
Stefan-Boltzmann constant (5.67 × 10 -8 W/m 2K4)
Atmospheric transmissivity ( - )
r Radiative fraction ( - )
-
8/19/2019 Radiacion por fuego
17/146
- 1 -
Chapter 1
INTRODUCTION
1.1 Context and Motivation
The dominant mechanism for the spread of fire within buildings is direct thermal radiation
from the existing flames (Karlsson & Quintiere, 2000). In order to determine if or when
certain objects adjacent to the fire may ignite or be damaged, one must be able to predict
the thermal radiation field surrounding the fire. This requires the radiant heat flux to be
determined at various points in space.
Being able to calculate the radiant heat flux from a fire provides a number of benefits.
These include:
• Prediction of if or when adjacent objects may ignite
• Prediction of extent of damage from fire
• Estimate safe separation distances between objects or buildings
• Estimate safe separation distances between a burning object and an escape route
• Prediction of activation of thermal detectors or sprinkler heads
• Helps to determine the total amount of heat transfer occurring between objects
• Prediction of failure of structural elements
Calculating the thermal radiation field surrounding a fire requires one or more equations to
be solved. This can either be performed manually or be programmed into a computer
model. This research uses the manual method, with the intent of providing
recommendations so that programming into a computer model can be achieved. Including
a thermal radiation algorithm into a computer model can provide fire engineers with a
useful tool to be used in analysis and design. In determining thermal radiation hazard, fire
engineers are interested in the maximum heat flux received by a differential area from a
given emitting source at a distance (He, 2001).
The specific motivation for this research is that a thermal radiation model is desired to be
input into an existing computer program named BRANZFIRE (Wade, 2008). This
-
8/19/2019 Radiacion por fuego
18/146
- 2 -
comprises part of a larger project being carried out between the Building Research
Association of New Zealand Ltd (BRANZ Ltd) and the University of Canterbury (UC).
The overall project is being funded by the Foundation for Research, Science and
Technology (FRST). The specific part of the project carried out in this research involves
evaluating the performance of a number of thermal radiation models under a range ofconditions. Sections 1.2-1.5 contain more information.
1.2 BRANZFIRE Software
A performance-based fire engineering design for a building typically requires some form
of quantitative analysis. There are a number of ways in which this can be carried out; one
of which is using a ‘zone model’. The term ‘zone model’ usually refers to a two-zone
model, based on the conceptual representation of the compartment fire process (Quintiere,2002). Here, the system is assumed to contain two distinct homogeneous gas layers (or
zones); a hot upper layer containing products of combustion and a relatively cool layer
beneath it. Conservation equations for mass and energy are solved numerically at each
time step for both the upper and lower zones. The flow of smoke and toxic products out
through compartment openings is also calculated. The fire itself is represented as a source
of energy and mass, which governs the amount of air entrainment that occurs into the
plume (Quintiere, 2002). The most important input parameters that must be specified by
the user of the software are the building geometry and a design fire. Design fires are
covered extensively in literature; a good overview is provided by Karlsson and Quintiere
(2000).
One such zone model that is commonly used by practising Fire Engineers in New Zealand
is BRANZFIRE (Wade, 2008). BRANZFIRE is a computer fire model which integrates a
flame spread and fire growth model for room lining materials with a multi-room zone
model (Wade, 2004). This research involves helping to develop the next version of
BRANZFIRE.
1.3 Probabilistic Design Tool
BRANZFIRE, like most other zone models at the time of writing, currently carries out its
modelling in a deterministic fashion. This means that a single set of outcomes is produced
-
8/19/2019 Radiacion por fuego
19/146
- 3 -
for a set of input parameters. There is, however, much interest in carrying out
probabilistic (or risk-based) analysis during fire engineering design. The Society of Fire
Protection Engineers (2005) defines Fire Risk Assessment as “a process for estimation and
evaluation of fire risk that addresses appropriate fire scenarios and their probabilities and
consequences, using one or more acceptability thresholds” which “develops the basis forfire risk management decisions.” Fire Risk Management is then defined as “the process of
deciding what should be done about the identified hazards, the exposed population, and
the foreseeable adverse outcomes. Fire risk management involves implementing a design
evaluated using fire risk assessment and managing an ongoing program (e.g., training,
maintenance) required to ensure that the adopted design continues to deliver the calculated
acceptable risk.” The implementation of risk assessment into computer fire modelling is
thought to become increasingly popular in the near future (Beyler, DiNenno, Carpenter, &
Watts Jr., 2008).
There is currently a research initiative between BRANZ Ltd and the University of
Canterbury which aims to include risk-based modelling in the forthcoming version of
BRANZFIRE. BRANZ Ltd (2007) describes the research initiative in detail. The
following is an excerpt from the document’s executive summary:
This research will develop a building fire design and analysis tool to simulate
building fire outcomes in a risk-descriptive format that will account for the
variability and uncertainty associated with the development of a fire, the nature
and arrangement of the building contents and the inherent reliability and
effectiveness of different fire safety features used to mitigate the risk of fire. The
intermediate outcome to which this research is directed is to help ensure that the
management of fire risks in buildings is cost-effective as well as socially and
politically acceptable, leading to innovation in construction, flexibility in design
and robustness in fire safety solutions.
Furthermore, BRANZ Ltd (2007) outlines that:
Our proposed design tool will use an existing mathematical fire model
(BRANZFIRE) for predicting the spread of smoke and development of hazardous
conditions in a building. The novel feature of the research will be extending
existing knowledge to develop a design tool that allows a wide range of fire
scenarios for a building to be simulated within the model, using specialised
-
8/19/2019 Radiacion por fuego
20/146
- 4 -
random sampling techniques for rare events. The tool would provide designers
with a means of conducting a sensitivity analysis by proposing input fire scenarios
and corresponding fire characteristics suitable as design assumptions for New
Zealand buildings… Results would be expressed as probability distributions. A
design tool with this capability has not been developed before in New Zealand orelsewhere.
By having a better understanding of the uncertainty involved in fire engineering designs, it
is hoped that the fire risks within buildings will be better managed and that there is more
robustness in fire safety solutions. This should lead to more transparency and confidence
in the level of fire safety provided in fire engineered designs.
Research for the BRANZFIRE project comprises staff from both BRANZ Ltd and UC and
students undertaking a masters or PhD in fire engineering at UC. Beginning in 2008, the
project is expected to take five years to complete and is funded by the Foundation for
Research, Science and Technology (FRST). The research described in this thesis makes
up a small part of the overall BRANZ Ltd – UC project, as outlined in Figure 1. Note that
the dates given in the diagram are given as an indication only and actual completion dates
may vary.
As can be seen in Figure 1, one element of the project is to develop an item to item fire
spread model that is to be included in BRANZFIRE. This technical basis for this item to
item fire spread model will come from work carried out in this thesis, coupled with
research performed by a PhD student. See section 1.5 for more details.
-
8/19/2019 Radiacion por fuego
21/146
- 5 -
Figure 1: Overview of development scope for risk-informed fire design tool (BRANZ Ltd, 2007). The
research outlined in this thesis comprises part of the highlighted section, Develop item-item fire spread
model
1.4 Thermal Radiation Overview
The three basic modes of heat transfer, namely conduction, convection and radiation, are
involved in almost all fire scenarios. It is observed that one mode dominates at different
stages of fire growth or in different locations. For example, conduction is of high
importance when trying to determine the expected temperature of a structural element
during a fire. It is radiation, however, that is the dominant mode of heat transfer for the
-
8/19/2019 Radiacion por fuego
22/146
- 6 -
spread of flames within compartments (Karlsson & Quintiere, 2000). It is the mechanism
by which items at a distance from a fire are heated up, which can lead to ignition without
direct flame contact. For these reasons, this thesis investigates thermal radiation only and
is not concerned with convective or conductive heat transfer.
Thermal radiation is the transfer of energy by electromagnetic waves. Radiant energy in
general can have a wide range of wavelengths; from radio waves with wavelengths of tens
of meters to cosmic rays with wavelengths less than 10 -14 m (Siegel & Howell, 1992).
Thermal radiation is detected as heat or light and occupies a narrow window in the
electromagnetic spectrum. This includes a small portion of the ultraviolet, all of the
visible light region and the infrared, corresponding to a wavelength range of
approximately 0.4 to 1000 m (Siegel & Howell, 1992). Thermal radiation is emitted
from tiny soot particles which are present in nearly all diffusion flames (Drysdale, 1999).
It is these soot particles which give the flame its characteristic yellow luminosity.
1.4.1 Emissive power and emissivity
The total emissive power of a flame is a function of temperature and wavelength, as
described by Planck’s Law, given in many radiation references such as Siegel and Howell
(1992). Here, the emissive power is for an ideal radiator, known as a ‘black body’.
However, real surfaces are not ideal radiators and therefore have an emissive power, E ,less than that for a black body. The fraction of radiation emitted in relation to the
maximum possible emission from a surface is called the emissivity, (Karlsson &
Quintiere, 2000). Therefore, a black body has an emissivity equal to unity.
In order to simplify thermal radiation calculations, the concept of a ‘grey body’ (or ‘ideal,
non-black body’) is introduced. For this, the emissivity is assumed to be independent of
wavelength (Drysdale, 1999). Furthermore, Kirchoff’s law states that the emissivity for a
surface is equal to its absorptivity. Karlsson and Quintiere (2000) advise that for
enclosure fire radiative exchanges, the grey body assumption is generally satisfactory.
The total radiation emitted, E , per unit area from a grey surface is given by Equation 1:4
f T E εσ = (1)
-
8/19/2019 Radiacion por fuego
23/146
- 7 -
Where is the emissivity, is the Stefan-Boltzmann constant (5.67 × 10 -8 W/m 2K4) and T f
represents the flame temperature (K). E can also be termed the emissive power of the
flame.
1.4.2 Configuration factors
The above equation can be used to calculate the radiative heat loss from a surface.
However, if one wishes to know the rate of heat transfer to a nearby object, the amount of
energy being radiated in that particular direction must be calculated. This can be done
using Equation 2, which introduces the concept of a configuration factor.4
12 f T F q σε =′′ (2)
Where q ′′ is the radiant heat flux (kW/m²) and F 12 is the configuration factor.
This factor takes into account the geometrical relationship between the emitter and the
receiver. Configuration factors (also known as shape or view factors) have a value
between zero and one. For example, when the receiver is very close to the flame and
oriented so that it is facing the fire, the configuration factor approaches one, as everything
viewed by the receiver is the flame (Iqbal & Salley, 2004). Davis and Bagster (1989)
explain that the configuration factor is dependent on three variables:
• The geometry of the emitter and receiver
• Whether the emitter and receiver can be ‘viewed’ by each other
• The direction of the exchange of thermal radiation
In this work, the configuration factor is determined for radiant energy exchange between a
finite surface (the flame) and a differential element at some distance from the flame. The
configuration factor is dependent on the dimensions of the finite surface and the distance
and angle between the emitter and target. Usually an assumption is made whereby the
flame is approximated as a simple shape such as a rectangle or cylinder, which enables
calculation of the configuration factor using established equations. Assuming that theflame takes on the shape of a cylinder or rectangle is far from an exact reproduction of the
observed geometry. However, due to the rapid fluctuation of the flame shape with time,
calculating an accurate configuration factor from the fire to a target would be an extremely
complicated and time intensive process. Section 2.3 provides more detail about methods
for determining configuration factors.
-
8/19/2019 Radiacion por fuego
24/146
- 8 -
1.5 Aim
This thesis aims to evaluate the performance of a number of thermal radiation models with
respect to item-to-item fire spread within a compartment. Models that are evaluated
include the point source and a number of cylindrical models. The effectiveness of these
models is tested for a variety of situations and recommendations are made to BRANZ Ltd
as to the suitability of different radiation models.
The scope of the analysis is limited to direct radiative impingement on a target and
therefore ignores contributions from the compartment surfaces and the hot upper gas layer.
Also, internal radiation within the flame is ignored. This radiative feedback to the fuel
surface helps to control the rate of burning during a fire. However, in the BRANZFIRE
model, a design fire is specified or chosen by the user and therefore the increased burning
rate due to radiation feedback within the flame is already accounted for.
Results from this research and analysis feed into work carried out by a PhD student. The
combined findings then make up the shaded part of Figure 1 ( Develop item-item fire
spread model ) which is a component of the overall BRANZ Ltd – UC project. The intent
is that the findings of this research will contribute to the development of the forthcoming
version of BRANZFIRE.
In order to evaluate the performance of the thermal radiation models being investigated,
the predictions made by these models are compared with experimental data. To obtain
this data, a comprehensive experimental programme has been undertaken. In these
experiments, measurements of radiant heat flux are taken at various positions surrounding
a gas burner. Pure propane gas is used as fuel to the burner, with mass flow controllers
regulating the gaseous flow. This allows fires of varying heat release rates to be specified.
Three different burner shapes were tested, with length to width aspect ratios of 1:1, 2:1
and 3:1. These aspect ratios are used to represent the likely shapes of furniture foundwithin common New Zealand buildings, such as chairs, sofas, tables and beds.
The measurements of radiant heat flux are taken using heat flux gauges, mounted on steel
frames, also called trolleys. Each frame holds four gauges, set at heights ranging from the
base of the flame to 1.5 m above the flame base. The trolleys wheels are set on rails,
-
8/19/2019 Radiacion por fuego
25/146
- 9 -
enabling the gauges to be moved to different distances from the fire. The gauges are also
able to be moved laterally across the face of the fire and can be oriented either vertically
or horizontally. Furthermore, the gas burner can be rotated to provide the heat flux gauges
with a different view of the fire.
By testing many different variables, not only is a vast array of experimental data collected,
but the limitations of some of the theoretical models are highlighted. These limitations are
usually due to large assumptions being made about the fire, which often do not hold true
in the real world. It is useful to know the situations in which the different radiation
models perform well and those in which the models become highly inaccurate. This
makes for a more comprehensive analysis as the different models can be compared under
different scenarios.
Following the comparison between the experimental results and predictions from the
radiation models, recommendations must be made to BRANZ Ltd. In terms of
implementing a thermal radiation model in the BRANZFIRE software, it is important that
the chosen model displays good accuracy coupled with ease of use. Therefore, this
research aims to not only evaluate the models, but recommend the most appropriate model
to be used by BRANZFIRE. This may involve finding a balance between model accuracy
and ease of programming and use.
The context of the investigation is with respect to compartment fires. As such, the fire
dimensions and heat release rates tested are restricted to those that are representative of
typical single-item compartment fires. In fact, the maximum fire size able to be tested is
limited by physical conditions such as the propane fuel supply and the size of the
laboratory. The investigation and subsequent findings therefore do not necessarily apply
to other fire scenarios; for example, large open liquid pool fires.
1.6 Outline of Thesis
This Master of Engineering in Fire Engineering thesis outlines the processes undertaken
and the results found for the evaluation of thermal radiation models for fire spread
between objects.
-
8/19/2019 Radiacion por fuego
26/146
- 10 -
Chapter 2 provides a review of the literature on thermal radiation modelling. Here, six
different radiation models are explained in detail, including all equations necessary to use
the models.
Chapter 3 outlines the methodology employed for the experimental programme of thisresearch. Details of the fire source, laboratory and instrumentation are given, in addition
to an outline of the experimental procedure. Techniques used to extract and analyse the
data are then explained. Finally, the method for determining the mean flame height is
given.
Chapter 4 provides all of the experimental results, accompanied by a discussion of these
results. Results are presented for heat release rate, radiant heat flux and mean flame
height.
Chapter 5 compares the experimental data with predictions made by the theoretical
radiation models. Here, the different models are closely scrutinised and their performance
under different conditions is discussed.
Chapter 6 summarises all of the findings from the research and makes conclusions as to
the appropriateness of the radiation models for different circumstances.
Recommendations are made to BRANZ Ltd about the most suitable model to be used for
the implementation of a radiation sub-model within BRANZFIRE.
Chapter 7 gives a list of references that have been used throughout this research. Finally,
the Appendices provide extra material not contained within the main body of the thesis.
-
8/19/2019 Radiacion por fuego
27/146
- 11 -
Chapter 2
LITERATURE REVIEW
The primary mechanism for injury or damage from large, open hydrocarbon fires is
thermal radiation (Beyler, 2002). With respect to fire radiation models, the majority of the
work has gone into predicting the thermal radiation from hydrocarbon pool fires (Beyler,
2002) and this will form the basis for this research.
2.1 Types of Model
Rew, Hulbert, and Deaves (1997) outline that two approaches are generally used todetermine the thermal radiation surrounding a fire. These are field models and semi-
empirical models. Firstly field models, commonly known as Computational Fluid
Dynamics (CFD) models, solve the Navier-Stokes equations for fluid flow across a vast
grid of cells, known as a mesh. Complex sub-models must be incorporated in order to
predict the chemical and physical processes occurring in a fire. The radiant heat transfer is
solved by means of an enthalpy conservation term that arises within the Navier-Stokes
equations (Cox & Kumar, 2002). The advantage of using field models for radiation
modelling is that they are capable of predicting a wide range of scenarios, provided thatthe input is correctly specified. However, there are disadvantages associated with field
models in that they require a lot of time and effort; both human (in terms of the input) and
computational (for solving the Navier-Stokes equations).
Semi-empirical models, on the other hand, are comparatively far easier to use and
understand. This means that they are more frequently used in risk assessments than field
models (Rew, et al., 1997). Semi-empirical models are designed to be simple to use and
therefore do not include complicated algorithms for the physical processes involved in
fires. As a result, a semi-empirical model designed to predict the radiant heat flux from a
fire is not designed to predict other phenomena. The correlations used in semi-empirical
models are derived from a wide range of experimental data and can provide more than
satisfactory results, provided they are used within their validation limits (Rew, et al.,
1997).
-
8/19/2019 Radiacion por fuego
28/146
- 12 -
This research deals exclusively with semi-empirical models and henceforth all radiation
models will be assumed to be semi-empirical.
2.2 General Approach to Thermal Radiation Modelling
Beyler (2002) describes the three major steps involved in estimating the thermal radiation
field surrounding a fire:
1. Determine the geometric characteristics of the fire, including the burning rate and
the physical dimensions of the fire. These dimensions are based on time-averaged
values.
2. Characterise the radiative properties of the fire. This involves the determination of
the average emissive power of the flames.
3. Calculate the incident radiant heat flux at the target location. For this to be carriedout, steps 1 and 2 must have been completed, as well as knowing the location,
geometry and orientation of the receiver.
The radiation models described in the following section use these three steps to varying
degrees of accuracy.
2.3 Common Radiation ModelsThe primary aim of radiation modelling usually is to calculate safe separation distances
between fire sources and potential targets that could be damaged or ignited by radiation
from the fire. These models range in the level of detail and rigour and some are more
suitable for certain applications than others. Some methods are most appropriate for crude
initial hazard assessments, while others are capable of more accurate predictions, although
more effort is required.
The following sections outline a number of thermal radiation models that are available in
the literature. Since the project is investigating compartment fires only, it is assumed that
the flames are not wind affected. There are thermal radiation models available in the
literature for wind-blown flames, such as the Mudan method (Mudan, 1984).
-
8/19/2019 Radiacion por fuego
29/146
- 13 -
2.3.1 Shokri and Beyler correlation
Shokri and Beyler (1989) developed a simple correlation based on experimental data from
large-scale pool fire experiments. This method calculates the radiant heat flux at ground
level as a function of the radial position of a vertical target. Note that the term ‘ground
level’ is loosely used to represent the height of the base of the fire. The heat flux received
by the target is given by Equation 3:59.1
4.15−
=′′ D L
q (3)
Where L is equal to the distance between the target and the centre of the fire (m) and D is
the fire diameter (m).
The correlation was derived for circular pool fires, however, for non-circular pools with a
length to width ratio of approximately one, an equivalent area circular source may be used
(Shokri & Beyler, 1989). The equivalent diameter is given by:
π A
D4= (4)
Where A is the cross-sectional surface area of the fire or pool (m).
The following assumptions apply for this method (Beyler, 1999)
• Pool is circular or nearly circular
• Target is vertical and located at ground level
Beyler (1999) lists some limitations with the model. They are as follows:
• The fuels used in the experiments that produced the correlation all produced
luminous flames. Therefore, the correlation may not be suitable for non-luminous
flames
• In the experiments, pool diameters of 1 to 50 m were used. It is reported (Beyler,
1999) that the correlation systematically over-predicted the results from the 50 mpool fire experiment. Therefore, the model should be used with great care for
larger pool diameters. For the smaller pools, the correlation yielded predictions
that were within ±100 % of the measured value
• The edge of the circular pool is at L/D equal to 0.5. Using Equation 3 above, this
yields a radiant heat flux of 46.3 kW/m², which is significantly less than values
-
8/19/2019 Radiacion por fuego
30/146
- 14 -
that were measured at that location during experimentation. It is recommended
that the correlation be limited to L/D values ranging between 0.7 and 15
Beyler (1999) then recommends that a safety factor of two should be applied to Equation 3
when being used for design purposes. However, if a realistic result is desired, no safetyfactor should be applied.
2.3.2 Point source model
The point source model (Modak, 1977) is the simplest configurational model of a radiant
source. The essence of the model is that radiation is assumed to emanate isotropically
from a single point source located at the centre of the flame, as shown in Figure 2a. The
relationship varies with the inverse square of the distance R from the source, as given bythe following equation:
24cos R
Qq r
π θ =′′ (5)
Where r Q is the total radiative energy output of the fire (kW), is the angle between the
normal to the target and the line of sight from the target to the point source location
(radians), and R equals the distance from the point source to the target (m).
The location of the theoretical point source of energy is at the centre of the fire at the mid-
height of the flame (see Figure 2b). The mean flame height, H , measured in m, is
calculated by the Heskestad correlation:
DQ H 02.123.0 52
−= (6)
Where Q is the heat release rate of the fire (kW).
The distance, R, from the point source location to the target location can be determined
using the Pythagorean Theorem, as given below for the given application:
22T H L R +=
(7)
Where H T is the height of the target relative to the height of the equivalent point source at
H /2 (m).
-
8/19/2019 Radiacion por fuego
31/146
- 15 -
Therefore, for a target located on the ground, H T = H /2. For a target at the mid-height of
the flame, H T = 0 .
As with the Shokri and Beyler correlation, an equivalent diameter can be used for non-
circular pools, given that the length to width ratio is near one. The effective diameter iscalculated using Equation 4. The total radiative energy output of the fire can be calculated
from Equation 8 below.
QQ r r χ = (8)
Where r is the radiative fraction.
(a) (b)
Figure 2: (a) Schematic of point source model (Karlsson & Quintiere, 2000), (b) Schematic and
notation for point source model (Beyler, 2002)
Generally, the radiative fraction, r , is dependent on the fuel type, flame size and flame
configuration. Its value can vary from approximately 0.15 for low-sooting fuels, such as
alcohol, to around 0.6 for high-sooting fuels, such as hydrocarbons (Iqbal & Salley, 2004).
The following assumptions apply for this method (Beyler, 1999):
• Pool is circular or nearly circular
• The point source configuration factor is used, as per Equation 5
Some limitations exist with the point source model. They are as follows:
• The point source model is a very simplistic model of a pool fire (Beyler, 1999)
-
8/19/2019 Radiacion por fuego
32/146
- 16 -
• The most important parameter in the model is the estimation of the radiative
fraction (Mudan, 1984) and therefore great care should be taken when choosing a
value for this. The radiative fraction is dependent on the fuel used
• The method is known to under-predict incident heat fluxes at locations close to the
fire (Drysdale, 1999). According to Iqbal and Salley (2004) this is primarilybecause the near-field radiation is greatly influenced by the flame size, shape, and
tilt as well as the relative orientation of the target
• The model performs poorly at heat fluxes at the target greater than 5 kW/m²,
indicating that it is not a good choice when ignition of combustibles is to be
considered (Beyler, 2002)
• The point source model is within 5 % of the measured incident heat flux when
L/D > 2.5 (Modak, 1977)
• The point source model “is a correct assumption at large distances from the fire”
(Beyler, 1999)
A safety factor of two is recommended for use with the point source method for design
considerations (Beyler, 1999), although it is recommended that this only applies for heat
fluxes less than 5 kW/m² as this is the recommended limit of the model. Like the Shokri
and Beyler correlation, no factor of safety should be applied if an accurate prediction is
desired. Beyler (1999) suggests that the point source model is the most appropriate
method for heat fluxes less than 5 kW/m².
Despite its simplicity, the point source model is often used for a range of applications.
One such example is for industrial flare design, where the model is seen to provide
adequate predictions of the thermal radiation field surrounding the flare (Oenbring &
Sifferman, 1980).
2.3.3 Shokri and Beyler detailed method
The methods outlined in sections 2.3.1 and 2.3.2 are known to be simple screening
methods (Beyler, 1999) and may not be appropriate if an accurate analysis is desired.
There are a number of more detailed methods available, one of which is presented by
Shokri and Beyler (1989). The basis of the model is to provide a simple yet realistic
model of the flame. To achieve this, the flame is assumed to be a cylindrical, black-body,
-
8/19/2019 Radiacion por fuego
33/146
- 17 -
homogeneous radiator with an average emissive power. It is assumed that thermal
radiation is emitted from the surface of the cylinder and that radiation from non-visible
gases is negligible (Iqbal & Salley, 2004). Like many fire radiation models, this method
was developed using pool fire radiation data.
Figure 3 provides a schematic and the nomenclature for the Shokri and Beyler detailed
method for both vertical and horizontal targets located at ground level. For targets above
ground level, the cylinder must be broken down into two individual cylinders, as shown in
Figure 4. In such instances, one cylinder represents the flame below the height of the
target, while the other represents the flame above the height of the target.
Figure 3: Cylindrical flame shape configuration factor geometry for vertical and horizontal targets atground level (Beyler, 2002)
Figure 4: Two-cylinder representations of the configuration factor for target above ground level
(Beyler, 2002)
-
8/19/2019 Radiacion por fuego
34/146
- 18 -
The incident radiative flux to a target outside the flame is given by Equation 9.
12 EF q =′′ (9)
The configuration factor is a function of the target location and the flame height and
diameter. F 12 always takes a value between zero and one, depending on these factors. For
non-circular fires, an effective diameter can be calculated using Equation 4. The flame
height can be determined using Equation 6.
Using the flame height and diameter, the configuration factors for horizontal ( F 12,H ) and
vertical ( F 12,V ) targets can be calculated using the Equations 10 and 11.
( )( )
( )( )
( )( )
( )( )1111
tan1
1
11
11tan
1
1
1
2
1
2,12
+−
−+
−
−
−
+−
−+
−
−
= −−
S A
S A
A
S A
S B
S B
B
S B
F H π π
(10)
( )( )
( )( )( )( )11
11tan
111
tan1
tan1 1
2
1
2
1,12 +−
−+
−+
+
−−
−= −−−
S AS A
AS
AhS S
S h
S
hS
F V π π π
(11)
Where:
D
H h
D
LS
S S
BS S h
A
2,
22
1,
21 222
==
+=
++=
(12)
The maximum configuration factor at a point, F 12,max , is given by the vector sum of the
horizontal and vertical components:
2,12
2,12max,12 V H F F F +=
(13)
Alternatively, Beyler (1999) provides five figures which display pre-calculated maximum
view factors for different ratios of flame height to radius. These may be useful when the
target is above ground level.
For vertically oriented targets located above ground level, Equation 11 must be applied for
both cylinders 1 and 2 (see Figure 4), yielding two configuration factors, F 12,V1 and F 12,V2 .
-
8/19/2019 Radiacion por fuego
35/146
- 19 -
The total configuration factor is given by the sum of the two individual configuration
factors.
2,121,12,12 V V V F F F += (14)
Horizontal targets, on the other hand, only require one equation as the target will only
receive thermal radiation from one of the two cylinders. This is because although the
target is infinitely thin, it only receives radiation on one of its faces. The user must decide
which surface is required; either the downwards-facing or upwards-facing surface. If the
thermal radiation to an upwards-facing surface, such as a desk top, is required then
Equation 10 should be employed using cylinder 2 in Figure 4. Conversely, if the user
wishes to calculate the radiant heat flux to a downwards-facing surface, such as the
underside of a table, cylinder 1 is treated as the sole emitter of radiation.
Shokri and Beyler (1989) explain it is important to note that the ‘effective’ emissive
power of the flame is defined only in terms of a homogeneous flame radiation model.
Rather than being the local emissive power measured at a specific point in space, it is
more of an averaged emissive power over the whole flame. As the model was developed
for pool fire scenarios, an expression for the ‘effective’ emissive power was formed in
terms of the effective pool diameter. It is:
( ) D E 00823.01058 −= (15)
Shokri and Beyler (1989) observed that the major uncertainty with their model is in the
definition of the emissive power and not in the view factor model. In fact, it was found
that for pool fires the cylindrical approximation of the flame is highly accurate at
predicting view factors over a wide range of conditions.
As with the previous two methods, this model assumes that the fire is circular or nearly
circular in shape. Comparison with experimental data suggests that the performance ofthe method is better at heat fluxes greater than 5 kW/m² at the target (Beyler, 1999).
Therefore, the main limitation to the model is that it should only be used when the radiant
heat flux to the target exceeds 5 kW/m². Again, a safety factor of two should be used for
design purposes (Shokri & Beyler, 1989).
-
8/19/2019 Radiacion por fuego
36/146
- 20 -
2.3.4 Mudan method
Sometimes known as the ‘Mudan and Croce method’, this model estimates thermal
radiation from pool fires for both wind-free and wind-blown flames. As this research
investigates the wind-free condition only, the models for wind-blown flames are omitted.
The radiative heat flux to a target is given by:
τ 12 EF q =′′ (16)
Where is the atmospheric transmissivity.
As with various other radiation models, this method centres around the assumption that
the flame is cylindrical in shape. Therefore, the flame height and diameter must be
determined. For noncircular fires, the effective diameter can be calculated using Equation
4, whilst the flame height correlation for this method is different from the previous
methods. Here, the correlation for mean visible height of turbulent diffusion flames,
developed by Thomas (1963), is used:61.0
42
′′= ∞
gD
m D H
a ρ (17)
Where ∞′′m is the mass burning rate per unit area (kg/m²s), a denotes the ambient air
density (kg/m³) and g is the acceleration due to gravity (9.81 m/s²).
The radiation exchange factor between the fire and a target outside the flames is dependent
on the flame shape (assumed to be cylindrical), the distance between the fire and the
target, and the relative orientation of the target. The maximum view factor at a point is
determined using Equations 10 – 13, given in section 2.3.3.
The effective emissive power, E , of the flame can be determined by the following
correlation:( ) ( )( )sDssD e E e E E −− −+= 1max (18)
Where E max is the equivalent black body emissive power (kW/m²), s is the extinction
coefficient (m -1) and E s represents the emissive power of smoke (kW/m²).
As shown in Equation 16, allowance must be given for atmospheric absorption and
scattering. This comes in the form of a transmissivity factor, . The main atmospheric
-
8/19/2019 Radiacion por fuego
37/146
- 21 -
constituents that are found to absorb thermal radiation are water vapour (H 2O) and carbon
dioxide (CO 2) (Mudan, 1984).
The following procedure, provided by Mudan (1984), is used to calculate the
transmissivity. Firstly, calculate the partial pressure of water vapour, p' w (atm), in theatmosphere:
−=′
aw T
RH p
53284114.14exp
100 (19)
Where RH indicates the relative humidity (%) and T a is the ambient temperature.
Next, determine the partial pressure path length parameter, pw (atm m) :
′
=a
f
ww T
T p p (20)
Where is the path length from the flame surface to the receiving target (m).
For the flame temperature and pw , determine the water vapour emissivity, w, using
emissivity plots given in Appendix A, Figure A 1.
Now the water vapour absorption coefficient, w, can be calculated from:45.0
=
f
aww T
T ε α (21)
The absorption by carbon dioxide is calculated in a similar fashion. Knowing that the
partial pressure of CO 2 remains relatively constant at about 3 × 10 -4 atm (Mudan, 1984)
and using Figure A 2 (Appendix A), the carbon dioxide absorption coefficient, w, is:65.0
=
f
acc T
T ε α (22)
Where c is the carbon dioxide emissivity.
Finally, the transmissivity can be determined from:
cw α α τ −−= 1 (23)
-
8/19/2019 Radiacion por fuego
38/146
- 22 -
Alternatively, the method can be simplified by assuming that = 1 (Beyler, 1999). This
eliminates the need for the user to consult plots, meaning that the method can be easily
programmed into a computer model or spreadsheet.
As with previous methods, the fire is assumed in to be circular or nearly circular in shape.Comparison with experimental data shows that the Mudan method is inherently
conservative for predicting radiant heat fluxes. Despite this, a safety factor of two should
still be applied when using the method for design purposes (Beyler, 1999).
2.3.5 Dayan and Tien method
A method presented by Dayan and Tien (1974) again approximates the flame as a
homogeneous cylinder of uniform temperature and other properties. Their methodcalculates the incident radiant heat flux from the flame to a target element, dA, with a unit
normal vector k w jviun ++= . Figure 5 depicts Dayan and Tien’s model.
Figure 5: Schematic of radiation exchange between a target element, dA, and a homogeneous
cylindrical flame (Karlsson & Quintiere, 2000)
The heat flux to the target is given by:( )3214 F F F T q f ++=′′ σε (24)
Where: µ ε 7.01 −−= e (25)
-
8/19/2019 Radiacion por fuego
39/146
- 23 -
β κ µ
sin2r = (26)
220π θ
β +
= (27)
( )002
1 2sin24θ θ π
π +−
= Lr u
F (28)
( )002 2sin22θ θ π
π +−
= Lr v
F (29)
02
3 cos θ π
= Lr w
F (30)
Where is the effective flame absorption coefficient (m -1), is the mean value of
(radians), 0 represents the angle between the z axis and the line of sight from the target to
the centre-top of the cylinder (radians), while u, v and w are the components of n in the i,
j and k directions, respectively (see Figure 5).
This method can be employed for predicting the radiant heat flux to targets located both at
ground level and at elevated positions (Dayan & Tien, 1974). For targets above ground
level, the cylinder which approximates the fire must be divided into two cylinders, in a
similar fashion to the Shokri and Beyler detailed method (see section 2.3.3).
The approximations provided in Equations 28 to 30 are deemed to be applicable for
L / r 3 (Dayan & Tien, 1974), where r is the fire radius (m). An investigation by He
(2001) found that the Dayan and Tien method is not as accurate as the Shokri and Beyler
detailed method when predicting the shape factor in the x direction. This is because
Equations 28 to 30 are seen to be approximations of the exact configuration factors
between a cylinder and a differential element. Equations 10 and 11 represent the exact
solutions in the y and z planes. The x component was assumed to be negligible. However,
the advantage of Dayan and Tien’s method over that of Shokri and Beyler is the relativesimplicity in its mathematical expressions.
-
8/19/2019 Radiacion por fuego
40/146
- 24 -
2.3.6 Rectangular planar model
In addition to the existing models and methods available in the literature, an attempt was
made to develop an original model. The objective for this new model was to move away
from the common assumption of a cylindrical flame shape. This is because the cylindrical
models were mostly developed for liquid pool fires, which are often contained in a tank of
circular cross section. However, compartment fires usually relate to burning solids such
as furniture, which typically have a rectangular cross section as opposed to circular. A
rectangular based model allows calculation of radiant heat flux to targets from fires that
are far from circular. In fact, in this research, fires are tested which have a length to width
aspect ratio of 3:1.
The basis for the determination of the shape factor for this model is that the flame can be
approximated as two perpendicular intersecting planes (see Figure 6a). The line of
intersection between these two planes extends vertically from the centre of the fire. The
two intersecting planes represent the centreline planes of a rectangular cuboid, as shown in
Figure 6b. The intent is that this represents a more universal flame shape assumption than
the cylindrical assumption. Plane has a length, l f (m), equal to the base length of the fire
source while Plane has a width, w f (m), equal to the base width of the fire source. Both
planes have a height, H , equivalent to the mean flame height.
(a) (b)Figure 6: (a) The rectangular planar model is made up of two perpendicular intersecting planes, (b)
The planes intersect at the centre of the fire, which can be approximated as a rectangular cuboid
The rectangular flame shape assumption for this method is not an original concept.
Drysdale (1999) describes that “the flame can be approximated by a simple geometric
w
l
H
Plane
Plane
-
8/19/2019 Radiacion por fuego
41/146
- 25 -
shape, such as a rectangle of height between 1.5 and 2 times the fuel bed diameter…”
Assuming a flame height based on the fire diameter is a very simple but crude method. A
method presented by Seeger (1974) also uses a rectangular surface assumption; however,
just one plane is used. The method is only valid for vertical targets and the plane must be
parallel to the target. Seeger’s method was derived as a simplification to the cylindricalassumption for circular pool fires. Its applicability to fires of other shapes, such as
rectangular cuboids, was not tested. This type of method is also known as an ‘equivalent
radiator’ model (Crocker & Napier, 1986). Robertson (1976) suggests that for circular
tank fires the equivalent radiator be a rectangle of width D and height 2 D . For fires that
approximate a rectangular cuboid (as the one depicted in Figure 6b), Robertson (1976)
recommends a similar method to that outlined above; with a rectangle of length equal to
the horizontal flame length and height equal to the flame height. Only one rectangle is
specified.
The radiant heat flux received by a differential target from planes and is calculated
using Equation 2, reproduced below.4
12 f T F q σε =′′ (2)
Calculation of F 12 uses an existing formula for the configuration factor between a finite
rectangle and a differential element located at some distance from the rectangle (see
Figure 7). Given by Howell (2008), the formula allows for the differential element to be
oriented at any angle to the rectangle. This feature makes it suitable for use in this
research, as then it can be programmed into a computer model which is able to calculate
the configuration factor for any situation. The configuration factor from a finite rectangle
to a differential element can be calculated using Equation 31.
-
8/19/2019 Radiacion por fuego
42/146
- 26 -
Figure 7: Pictorial representation and notation for the configuration factor from a finite rectangle to a
differential element (Howell, 2008)
++
−+
++
−+×+×
=−
−−−
2
1
2
2
1
2
11
12
1tan
1
coscos1
tan1
coscoscostancostan
21
B
A
B
B A
B
A
A A B
F jk
ik ji
θ θ
θ θ θ θ
π (31)
Where:
cb
Bca
A == ; (32)
Figure 7 shows that the target must be directly in line with one corner of the finite
rectangle in a direction normal to the plane of the rectangle. This would appear to greatly
limit the utility of the formula as targets must always correspond to a corner. Clearly, in
fire situations a target will not always meet this criterion, as there would be potential
targets for ignition all over the room. In order to solve this problem, one must add or
subtract various components that make up the overall configuration factor. This is most
easily explained diagrammatically.
Figure 8 depicts just one of the fire planes for simplicity, that being Plane (labelled A2).
In the situation shown, when a normal is taken from the zy plane towards the target ( dA1),
the normal lies within the bounds of A2. At the point where the normal intersects A2,
horizontal and vertical lines are projected out to the boundaries of the rectangle. This
-
8/19/2019 Radiacion por fuego
43/146
- 27 -
creates four individual rectangles, each with a configuration factor to dA1 which can be
calculated using Equation 31. Each of the four individual configuration factors F 1 to F 4
contribute to the overall configuration factor, F . For the situation described by Figure 8,
the overall configuration factor is:
F = F 1 + F 2 + F 3 + F 4
Figure 8: Example situation where the normal of the rectangle to the target lies within the bounds of
the fire (area A 2). The fire must be divided into four individual rectangles and the configuration
factors for each added to achieve the overall configuration factor
A different situation is outlined in Figure 9. Here, when a normal is taken from the zy
plane towards the target, the normal does not actually intersect A2. To calculate the
overall configuration factor, the factors F 1 and F 2 must be added and then F 3 and F 4 must
be subtracted as these are not part of A2.F = F 1 + F 2 – F 3 – F 4
Figure 9: Example situation where the normal of the rectangle to the target lies outside of the bounds
of the fire (area A 2). The overall configuration factor is found by adding factors F 1 and F 2 then
subtracting factors F 3 and F 4
dA1
A2
F1
F2
F3
F4
y
z
y
z
dA1
A2
F1
F2F3
F4
-
8/19/2019 Radiacion por fuego
44/146
- 28 -
The same method can then be employed for calculating the configuration factor from
Plane to the target. The only requirement is that the definition of i,j,k is consistent with
that for Plane . The overall shape factor is found by addition of the two components F
and F .
β α F F F += (33)
Where F is the configuration factor from Plane to the target and F is the configuration
factor from Plane to the target.
The flame emissivity is calculated using the following equation: De κ ε −−= 1 (34)
As this is a new model, there is currently no validatory data for the rectangular planar
model.
2.4 Variables for Models
As described in section 2.3, some of the models require the knowledge of a number of
variables. These variables are essential inputs to the models; therefore, care should be
taken in selecting appropriate values. The variables required, along with a review of the
literature as to appropriate values, are described in the following sections.
2.4.1 Radiative fraction
In order to accurately predict the radiant heat flux to a target, one must determine the
fraction of total combustion energy that results in thermal radiation. Known as the
radiative fraction, r , it is a function of the efficiency of combustion and the formation of
soot, as well as the heat that is convected away from the fire. Markstein (1976) found that
the radiative fraction is independent of the heat release rate of the fire. Table 3-11.12 of
Beyler (2002) reports that the radiative fraction of propane gas is between 0.30 and 0.32.
In the thermal radiation modelling presented in this thesis, the radiative fraction is treated
as a variable. It should be noted that Sivathanu and Faeth (1990) determined a radiative
fraction value of 0.28 for propane gas; however, this was for a small and possible laminar
-
8/19/2019 Radiacion por fuego
45/146
- 29 -
flame. The flames that are of interest in this research are most definitely turbulent;
therefore, Sivathanu and Faeth’s value has little applicability to this situat